Applications of software radio for hand gesture recognition by using long training symbols

This paper presents a novel application of software radio (e.g., Sora) for hand gesture recognition by using long training symbols. The type of gesture performed between the transmitter and receiver can have significant effects on the received wireless signals (e.g., IEEE 802.11a). Since all wireless signal processing functions could be done completely in software, we can easily capture the two long training symbols from each wireless frame in Sora. Then we extract the frequency offset and channel estimation parameters as representative features for gesture classification. We classify the gestures using Support Vector Machine classifier. The test results show that a set of eight static gestures can be identified and classified. It gives an average accuracy of 95% and illustrates the potential of this novel gesture recognition method for smart home and human computer interaction purposes.